libpcap之零拷贝mmap

一、用户空间

在通过socket(AF_PACKET,…)创建fd后，进行接收队列的建立

//pcap-linux.c
static int
pcap_activate_linux(pcap_t *handle) {
...
	ret = setup_mmapped(handle, &status);
...
}

1.1 设置默认ring bufer size

static int
setup_mmapped(pcap_t *handle, int *status)
{
	...
	//1.
	if (handle->opt.buffer_size == 0) {
		/* by default request 2M for the ring buffer */
		handle->opt.buffer_size = 2*1024*1024;
	}
	...
}

1.2 确定tpacket版本

不同的版本格式略有不同。

static int
setup_mmapped(pcap_t *handle, int *status)
{
	...
	//1.
	...
	//2.
	ret = prepare_tpacket_socket(handle);
	...
}

1.3 创建ring

static int
setup_mmapped(pcap_t *handle, int *status)
{
	...
	//1.
	...
	//2.
	...
	//3.
	ret = create_ring(handle, status);
	...
}

计算block_size,frame_size等

以TPACKET3为例

static int
create_ring(pcap_t *handle, int *status)
{
	struct pcap_linux *handlep = handle->priv;
	unsigned i, j, frames_per_block;
...
	struct tpacket_req3 req;
...
	socklen_t len;
	unsigned int sk_type, tp_reserve, maclen, tp_hdrlen, netoff, macoff;
	unsigned int frame_size;
...

	switch (handlep->tp_version) {

	case TPACKET_V2:
	...
		break;

	case TPACKET_V3:
		req.tp_frame_size = MAXIMUM_SNAPLEN;
		req.tp_frame_nr = (handle->opt.buffer_size + req.tp_frame_size - 1)/req.tp_frame_size;
		break;
	default:
		...
		*status = PCAP_ERROR;
		return -1;
	}
	
	req.tp_block_size = getpagesize();
	while (req.tp_block_size < req.tp_frame_size)
		req.tp_block_size <<= 1;

	...

请求内核创建接收队列ring buffer

	...
frames_per_block = req.tp_block_size/req.tp_frame_size;
...
	/* ask the kernel to create the ring */
retry:
	req.tp_block_nr = req.tp_frame_nr / frames_per_block;

	/* req.tp_frame_nr is requested to match frames_per_block*req.tp_block_nr */
	req.tp_frame_nr = req.tp_block_nr * frames_per_block;
...
	if (setsockopt(handle->fd, SOL_PACKET, PACKET_RX_RING,
					(void *) &req, sizeof(req))) {
		...
	}

将内核创建的ring buffer映射到用户空间

	/* memory map the rx ring */
	handlep->mmapbuflen = req.tp_block_nr * req.tp_block_size;
	handlep->mmapbuf = mmap(0, handlep->mmapbuflen,
	    PROT_READ|PROT_WRITE, MAP_SHARED, handle->fd, 0);
	...

初始化用户态的映射空间头

...
/* fill the header ring with proper frame ptr*/
	handle->offset = 0;
	for (i=0; i<req.tp_block_nr; ++i) {
		u_char *base = &handlep->mmapbuf[i*req.tp_block_size];
		for (j=0; j<frames_per_block; ++j, ++handle->offset) {
			RING_GET_CURRENT_FRAME(handle) = base;
			base += req.tp_frame_size;
		}
	}

	handle->bufsize = req.tp_frame_size;
	handle->offset = 0;
...

二、内核空间

2.1 创建 rx ring

// net/packet/af_packet.c
static int
packet_setsockopt(struct socket *sock, int level, int optname, char __user *optval, unsigned int optlen)
{
	struct sock *sk = sock->sk;
	struct packet_sock *po = pkt_sk(sk);
	int ret;

	if (level != SOL_PACKET)
		return -ENOPROTOOPT;

	switch (optname) {
	...
	case PACKET_RX_RING:
	case PACKET_TX_RING:
	{
		union tpacket_req_u req_u;
		int len;

		lock_sock(sk);
		switch (po->tp_version) {
		case TPACKET_V1:
		case TPACKET_V2:
			len = sizeof(req_u.req);
			break;
		case TPACKET_V3:
		default:
			len = sizeof(req_u.req3);
			break;
		}
		if (optlen < len) {
			ret = -EINVAL;
		} else {
			if (copy_from_user(&req_u.req, optval, len))
				ret = -EFAULT;
			else
				ret = packet_set_ring(sk, &req_u, 0,
						    optname == PACKET_TX_RING);
		}
		release_sock(sk);
		return ret;
	}
	
	default:
		return -ENOPROTOOPT;
	}
	...
}

分配空间

static int packet_set_ring(struct sock *sk, union tpacket_req_u *req_u,
		int closing, int tx_ring)
{
	struct pgv *pg_vec = NULL;
	struct packet_sock *po = pkt_sk(sk);
	...
	int was_running, order = 0;
	...
	/* Added to avoid minimal code churn */
	struct tpacket_req *req = &req_u->req;

	...
	order = get_order(req->tp_block_size);
	pg_vec = alloc_pg_vec(req, order);

初始化

...
switch (po->tp_version) {
		case TPACKET_V3:
			/* Block transmit is not supported yet */
			if (!tx_ring) {
				init_prb_bdqc(po, rb, pg_vec, req_u);
			} else {
				...
			}
			break;
		default:
			break;
		}

将新空间挂入队列上

mutex_lock(&po->pg_vec_lock);
	if (closing || atomic_read(&po->mapped) == 0) {
		err = 0;
		spin_lock_bh(&rb_queue->lock);
		swap(rb->pg_vec, pg_vec);
		rb->frame_max = (req->tp_frame_nr - 1);
		rb->head = 0;
		rb->frame_size = req->tp_frame_size;
		spin_unlock_bh(&rb_queue->lock);

		swap(rb->pg_vec_order, order);
		swap(rb->pg_vec_len, req->tp_block_nr);

		rb->pg_vec_pages = req->tp_block_size/PAGE_SIZE;
		po->prot_hook.func = (po->rx_ring.pg_vec) ?
						tpacket_rcv : packet_rcv;
		skb_queue_purge(rb_queue);
		if (atomic_read(&po->mapped))
			pr_err("packet_mmap: vma is busy: %d\n",
			       atomic_read(&po->mapped));
	}
	mutex_unlock(&po->pg_vec_lock);

2.2 将队列空间映射到用户空间

static int packet_mmap(struct file *file, struct socket *sock,
		struct vm_area_struct *vma)
{
	struct sock *sk = sock->sk;
	struct packet_sock *po = pkt_sk(sk);
	unsigned long size, expected_size;
	struct packet_ring_buffer *rb;
	unsigned long start;
	int err = -EINVAL;
	int i;

	...

	start = vma->vm_start;
	for (rb = &po->rx_ring; rb <= &po->tx_ring; rb++) {
		if (rb->pg_vec == NULL)
			continue;

		for (i = 0; i < rb->pg_vec_len; i++) {
			struct page *page;
			void *kaddr = rb->pg_vec[i].buffer;
			int pg_num;

			for (pg_num = 0; pg_num < rb->pg_vec_pages; pg_num++) {
				page = pgv_to_page(kaddr);
				err = vm_insert_page(vma, start, page);
				if (unlikely(err))
					goto out;
				start += PAGE_SIZE;
				kaddr += PAGE_SIZE;
			}
		}
	}

	atomic_inc(&po->mapped);
	vma->vm_ops = &packet_mmap_ops;
	err = 0;

out:
	mutex_unlock(&po->pg_vec_lock);
	return err;
}